MobileFaceNet_Tutorial_Pytorch
我训练的:25.73上,加了loss增强后,不加貌似没有效果
0.9351_1.68_0.963_1.56_0.9957_1.47_model_1.0e-04_32
0.9356_1.69_0.963_1.56_0.9953_1.47_model_1.0e-05_6
官方:
Eval Type | Score | LFW | AgeDB-30 | CFP-FP |
Flip | L2 | 99.52 | 96.3 | 92.93 |
Flip | Cos | 99.5 | 96.18 | 92.84 |
UnFlip | L2 | 99.45 | 95.63 | 93.1 |
UnFlip | Cos | 99.45 | 95.65 | 93.1 |
512维
criterion = torch.nn.CrossEntropyLoss().to(device)
optimizer_ft = optim.SGD([
{'params': model.parameters(), 'weight_decay': 5e-4},
{'params': margin.parameters(), 'weight_decay': 5e-4}], lr=0.0001, momentum=0.9, nesterov=True)
exp_lr_scheduler = lr_scheduler.MultiStepLR(optimizer_ft, milestones=[6, 8, 10], gamma=0.3)